Se ha denunciado esta presentación.
Utilizamos tu perfil de LinkedIn y tus datos de actividad para personalizar los anuncios y mostrarte publicidad más relevante. Puedes cambiar tus preferencias de publicidad en cualquier momento.

The Role of Scientific Method in Software Development

2.713 visualizaciones

Publicado el

Talk originally prepared for my Honorary Doctorate by Blekinge Institute of Technology in 2009 and reused for EASE Conference keynote in 2010

Publicado en: Tecnología, Empresariales
  • Sé el primero en comentar

The Role of Scientific Method in Software Development

  1. 1. The Role of Scientific Method in Software Development Natalia Juristo Universidad Politécnica de Madrid Spain
  2. 2. Content 1. 2. 3. 4. 5. State of Software Development The Scientific Method Experimental Software Engineering ESE Brief History The SE Lab
  3. 3. State of Software Development
  4. 4. Some Data on Sw Development  16.3% of software projects are successful The project is completed on time and within budget, and has all the features and functions specified at the start  52.7% of software projects cost more, take longer or do less The project is completed and operational, but it cost more than budgeted (189% more), took longer than estimated and offers fewer features and functions than originally specified (42%)  31% are cancelled The project is called off at some point during development before the system is put into operation
  5. 5. SE Knowledge     Do we know why some projects are successful and others are cancelled? Can we predict which project will fail and which one will not ? Do we know why some products (requirements, design, code) are obtained with higher quality than others? Can we predict it ? Are we able to predict anything about software development?
  6. 6. State of Software Development  State of software development knowledge    Today the results of applying software development methods are unpredictable There is no evidence to support most of the beliefs on which software systems development is based State of software development practice  Method selection for and decision making on software production is based on suppositions and subjective opinions   When, by chance (or thanks to practitioners’ personal and nontransferable know-how), the right methods are used, the software construction projects run smoothly and output the desired product When the wrong methods are applied, the project develops haphazardly and the output product tends to be of poor quality
  7. 7. From Craft to Engineering  SE must lay aside perceptions, bias and marketspeak to provide fair and impartial analysis and information  Decisions should be based on scientific and reasoned foundations. The gurus’ claims should not be docilely accepted
  8. 8. Other Engineering Disciplines  All engineering disciplines have taken a similar step that has released them from the craft status “The information used to build artefacts goes through stages, ranging from beliefs, speculations and lucky guesses to scientific knowledge by means of which an engineering discipline achieves predictable results” Mary Shaw Prospects for an engineering discipline of software IEEE Software 1990
  9. 9. Perceptions vs. Objectivity  Engineering disciplines are grounded on objectivity  Objective data are helpful for finding out more about reality, whereas subjective opinions can lead to mistaken perceptions of reality  Only by working with scientific evidences rather than assumptions will software development become a real engineering discipline
  10. 10. The Scientific Method
  11. 11. Science  Science is a process to understand the world  To understand a phenomenon, it is necessary to conduct methodical, thorough, meticulous and objective study
  12. 12. In Search of Knowledege Science is a way of thinking, more than a set of facts Carl Sagan 1934 – 1996
  13. 13. Science  Science seeks explanations about    how the world works why the world works as we perceive Such explanations are known as laws or theories
  14. 14. Scientific Laws    Are patterns of behaviour Describe cause-effect relationships Explain   why some events are related how the mechanism linking the events behaves
  15. 15. Scientific Law  The concept of law took a long time to come to fruition in human thought  Aristotle² discussed the motion of material bodies, but his natural laws of motion were just descriptions of how the final causes were supposed to behave “A stone falls because the “natural place” for heavy objects is the centre of the Earth”  Appreciate the difference on how Galileo studied that same topic Galileo³ dropped balls of the same material, but different masses, from the Leaning Tower of Pisa and observed that their time of descent was independent of their mass. This was contrary to what Aristotle had taught: that heavy objects fall faster than lighter ones, in direct proportion to weight ² 384BC – 322BC ³ 1564 - 1642
  16. 16. Scientific Law  Newton’s gravitation and motion laws ushered in the scientific era  Newton's postulate of an invisible force able to act over vast distances led to him being criticized for introducing "occult agencies" into science  We cannot perceive laws directly through our senses  1643-1727 Anyone can see an apple fall, but Newton’s inverse-square law of gravitation only becomes apparent through special systematic measurements
  17. 17. Causal Understanding   Gravity is the explanation of why bodies fall The relationships between the two bodies: the one falling and the one attracting it
  18. 18. Falsification   Falsification is the refutation of a hypothesis or theory by empirical evidence Karl Popper:    Universal hypotheses are falsifiable, but not verifiable “All swans are white” can be accepted as a tentative hypothesis, until the first non-white swan is sighted The longer a (falsifiable) hypothesis resists falsification, the more reliable it is seen 1902 - 1994
  19. 19. Knowledge can Always Improve  Einstein’s relativity revised Newton’s theory of gravity  With better predictions under new conditions (light speed, etc.) that Newton’s gravitation Law did not consider  Deeper understanding   In a way gravity does not exist. Gravity is a subproduct of the spacetime warping produced by heavy objects Newton’s gravitation law are still used today for nonrelativistic gravitational calculations
  20. 20. From Nexus to Laws  We cannot perceive laws directly through our senses  Two activities are necessary   Systematic objective observation Inference of links between cause & effect
  21. 21. The Scientific Method  Collection of Empirical Data   Theoretical Interpretation of Data   Systematic observation to appreciate the nexus Form a hypothesis (right or wrong) about the mechanisms relating the events Collection of Empirical Data  Hypothesis need to be tested against reality to know if they are true or not
  22. 22. What Makes a Method Scientific?  Exploration of causality  Objective observation  Methodic variation and control of variables  Repeatability
  23. 23. Objective Observation When you can measure what you are speaking about and express it in numbers, you know something about it; but when you cannot measure it, your knowledge is of a meager and unsatisfactory kind; it may be the beginning of knowledge, but you have scarcely in your thoughts advanced to the state of Science Lord Kelvin 1824 - 1907
  24. 24. Prescientific Stages  Before existing scientific knowledge humans are able to build interesting artifacts  Building does not mean understanding the mechanisms
  25. 25. Building without Understanding
  26. 26. 1st Step: Objective Observation Empirical Information Prediction Theoretical Interpretation Mechanism Understanding
  27. 27. Predicting without Understanding
  28. 28. Data Theory  The accumulation of data, without inferential leap to the theory, never reaches a mechanism understanding  In the absence of empirical data, theories are moving in the realm of speculation
  29. 29. 2nd Step: Objective Observation Empirical Information Prediction Theoretical Interpretation Mechanism Understanding
  30. 30. Wrong Explanation  Geocentric Theory  Systematic observation  Two common observations were believed to support the idea that the Earth is in the center of the Universe    The perception that as the Earth is solid and stable it is not moving but is at rest Hypothesis of explanation   The stars, sun, and planets appear to revolve around the Earth each day, with the stars circling around the pole and those stars nearer the equator rising and setting each day and circling back to their rising point Ptolemaic Planets System Similar ideas were held in ancient China
  31. 31. Wrong but Useful  Even shallow causal understanding is much better than nothing  Ptolemaic Planets (wrong explanation based on data) were useful for prediction  Antikythera mechanism (an ancient mechanical calculator built about 150–100 BC) was designed to calculate astronomical positions based on Ptolemaic Planet System  Predicted solar eclipses  Calculated the timing of the Ancient Olympic Games  …
  32. 32. Scientific Method: A Definition  A body of techniques for investigating phenomena and acquiring new knowledge  To be termed scientific, a method of inquiry must be based on gathering observable, empirical and measurable evidence subject to specific principles of reasoning
  33. 33. Scientific Method: Another View   A rigorous process for properly developing and evaluating explanations for observable phenomena based on reliable empirical evidence and neutral, unbiased independent verification Not based on arguments from authority or popular preferences
  34. 34. Experimental Software Engineering
  35. 35. Experimental SE  Brings the scientific method through the experimental paradigm into SE  Generate scientific statements about software building through experimentation
  36. 36. Why Experimenting in SE?  Too many methods and tools exist to allow each developer or software industry to determine the best choice of tools and methods by trial and error   Yet this choice is obviously important A general feeling of unease with the “advocacy style” of software research  The “Try it, you’ll like it”-method is not feasible anymore: too many alternatives to try
  37. 37. Claims for SE Experimentation: Old  Did Anyone Study Any real Programmers?   Experimentation can help build a reliable base of knowledge and thus reduce uncertainty about which theories, methods, and tools are adequate   Bill Curtis, Paper tile at Empirical Studies of Programmers Workshop, 1986 Tichy, Computer, 1998 Experimentation in SE is necessary, common wisdom, intuition, speculation, and proofs of concepts are not reliable sources of credible knowledge  Vic Basili, TSE, 1999
  38. 38. Claims for SE Experimentation: New  The actual software construction isn’t necessarily experimental, but its conception is. An this is where our focus ought to be. It’s where our focus always ought to have bee   Tom de Marco, IEEE Software, 2009 The empirical side of software engineering should become a full member of empirical sciences  Bertrand Mayer, Communications ACM, 2009
  39. 39. Experimentation to the Rescue!  Experiments determine whether claimed differences among alternative SE techniques are actually observable   For instance with respect to quality, cost, maintainability of the software… Experimentation is maturing    Fraction of empirical studies is rising More laboratories, more sophistication More and more papers contain empirical validation
  40. 40. Gaining SE Scientific Knowledge  Identify and understand   the variables that play a role in software development the connections between variables  Learn cause-effect relationships between the development process and the obtained products  Establish laws and theories about software construction that explain development behaviour
  41. 41. What can we hope?      Knowledge about which methods work better than others, under what circumstances, and why Trustworthy theoretical models that explain and predict Professional developers who can judge the validity of empirical studies... … who can ask meaningful questions about experiments and theories and understand the answers New methods and tools that provably simplify the programmers work and we know why Experimentalism is central to Software Research!
  42. 42. ESE Brief History
  43. 43. ESE: Origins  The origins of ESE can be traced back to Weinberg and Tichy’s work in the 1970s advocating scientific approaches that encouraged careful observation and evaluation of software development methods  The first experiments were not run until the early 1980s by Victor Basili’s group at the University of Maryland with NASA’s Software Engineering Laboratory  Since then the use of experiments to examine the applicability of SE technologies has gradually gained in importance as a research methodology
  44. 44. ESE At Present Journals & Conferences  A specialized journal and conference have been founded over recent years    Empirical Software Engineering Journal Empirical SE and Measurement Conference Experiments are reported in leading SE journals and conferences  For example, as many as 93 experiments were published in journals and conferences from 1993 to 2003
  45. 45. ESE At Present Textbooks  There are also a couple of books that explain how to apply experimental tests to software development  Experimentation in SE: An Introduction Wohlin, Runeson, Höst, Ohlsson, Regnell, Wesslén 2000  Basics of SE Experimentation Juristo & Moreno 2001
  46. 46. The SE Lab
  47. 47. What makes a Method Scientific?  Exploration of causality  Objective observation  Methodic variation and control of variables  Repeatability
  48. 48. Experiment  Experiment    Models key characteristics of a reality in a controlled environment and manipulating them iteratively to investigate the impact of such variations and get a better understanding of a phenomenon Formal, rigorous and controlled investigation in which the variables under study are given different values to find out what effect each value has The properties of a complex system are explained by analysing the behaviour of its parts
  49. 49. The Laboratory  Laboratory   Chemistry laboratory    Simplified and controllable reality where the phenomenon under study can be manipulated and studied Flasks and pipettes where temperatures and pressures are controlled Real world: Real substances with temperature and pressures Economics laboratory   Sets of individuals playing games to earn toy benefits Real-world: Markets (built of thousands of agents) where real rewards are pursued
  50. 50. SE Experiment   Development decomposed into its parts Manipulated variables     Techniques (design, testing, etc.), Developers (experience, knowledge, etc.) … Impacts investigated   Effectiveness, efficiency, productivity, quality Depending on the experiment, these response variables can be instantiated as, for example, number of detected defects, number of code lines, etc.
  51. 51. SE Laboratory     Students rather than professionals Toy software rather than real systems Exercises rather than real projects Academy or a tutorial in industry rather than real life in industry  Phases, techniques rather than project  …
  52. 52. Perceptions vs. Facts: Example   An experiment in our V&V course which studies techniques effectiveness Are perceptions reliable?  Perception    Facts    Students like much more functional testing technique They hate reading technique Functional and structural are equally effective Reading usually are more effective Tastes, previous knowledge, easiness, laziness are influencing their perceptions
  53. 53. Other Examples  Experiments falsify hypotheses   E.g. the assumption of failure independence in multiversion programming John Knight and Nancy Leveson (1986) -- An experimental evaluation of the assumption of independence in multiversion programming. IEEE TSE
  54. 54. From Lab to Clinic  Any lab finding has problems with representativeness of reality  Different levels of experimental studies    In vitro experiments In vivo experiments Clinical trials
  55. 55. THE ROLE OF SCIENTIFIC METHOD IN SOFTWARE DEVELOPMENT Natalia Juristo Universidad Politécnica de Madrid Spain
  56. 56. There is a lot more to say about this...  Lakatos’ “Research programs”    Kuhn’s “Paradigms”     Theory needs time and space to develop Falsification attempts should be delayed until it is well developed Science does not follow the falsification path Instead, there are “revolutions”, in which an established theory is replaced by a different one Between revolutions, there is “normal science” (extension, strengthening, application, rounding out of accepted theory.) Feyerabend’s “Anarchy”    From a historic perspective, there is no “scientific method” “anything goes” to find new theories The scientist must be competent, informed and follow scientific standards.
  57. 57. Prediction vs. Understanding Empirical Information Prediction Scientific Knowledege Mechanism Understanding